Rating Scale Use by Children With Disabilities on a Self-Report of Everyday Activities

2047

ORIGINAL ARTICLE

Rating Scale Use by Children With Disabilities on a Self-Report of Everyday Activities Jessica M. Kramer, PhD, OTR/L, Everett V. Smith Jr, PhD, Gary Kielhofner, DrPH, OTR ABSTRACT. Kramer JM, Smith EV Jr, Kielhofner G. Rating scale use by children with disabilities on a self-report of everyday activities. Arch Phys Med Rehabil 2009;90:2047-53. Objective: To examine whether children with disabilities interpret a self-report of perceived competence and importance of everyday activities in a consistent manner and use the rating scales as intended. If not, are differences in how children interpret the scale associated with personal or contextual variables? Design: Assessment and rating scale development using the Mixed Rasch Model. Setting: Rehabilitation clinics, schools, and research sites in the United States and Europe. Participants: Children (N⫽407) aged 6 to 17 years; all had a diagnosed disability or received occupational therapy services. Interventions: Not applicable. Main Outcome Measure: The Child Occupational Self Assessment; 25 items representing everyday activities and two 4-point scales: competence and importance. Results: For each scale, 2 groups of children were identified. Approximately 50% of the children used the rating scales as intended. The remainder used the scales as reversed 2-point scales; these children were younger and more likely to have an intellectual disability. Country and practice setting were also associated with rating scale use. All items but 1 had acceptable fit to the Rasch model, and groups of children differed in the relative competence and importance reported. Conclusions: Personal and contextual variables are associated with children with disabilities’ use of self-report rating scales. Younger children and children with intellectual disabilities use a modified response pattern. Key Words: Disabled children; Occupational therapy assessment. © 2009 by the American Congress of Rehabilitation Medicine EHABILITATION RESEARCH and practice increasingly R involve capturing children’s perspectives through the use of self-report measures. To self-report in a meaningful way,

an evaluation of oneself, and then identify the response that best matches that evaluation.6 Cognitive delays and other contextual factors can limit children’s understanding of the meaning of each item and rating category, thereby making the self-report process less valid.7,8 Determining how children with disabilities respond to self-reports enables researchers and practitioners to administer and interpret self-reports in the most appropriate manner. The COSA9 is designed to capture children’s perceptions of their competence for performing everyday activities and the importance attached to those activities. Everyday activities are defined as frequently performed, goal-directed actions that children execute in order to take care of themselves, interact with others, learn, and play.10,11 Modifications can be made during administration to ensure children can access and understand the assessment. Children’s responses are used to identify activities that should be addressed in rehabilitation or that can be used to motivate and engage the children. A series of Rasch measurement studies demonstrates that the COSA has good construct and internal validity.12-14 Analyses support that items measure 2 unidimensional constructs as all competence scale items have acceptable fit, and all but 1 importance scale items have acceptable fit.14 However, statistics revealed that not all children responded in a reliable manner and that fit was associated with personal and contextual variables including age, presence of intellectual disability, use of modifications, and country of residence.14 It is possible that children with these personal and contextual variables represent a separate population (ie, latent class) of respondents that professionals should be aware of when administering the COSA. Further, it is not known whether these children had difficulty interpreting the items, rating scale, or both. Therefore, this study asked, “Do all children in the current sample interpret items and use the rating scales in a consistent and expected manner?” (ie, do all children come from 1 latent class for which a Rasch model is appropriate), and if not, “Are different latent classes distinguished by personal and contextual variables?” METHODS

1-5

children must interpret a question, compare that question with

From the Department of Occupational Therapy, Boston University, Boston, MA (Kramer); Departments of Occupational Therapy (Kielhofner) and Educational Psychology (Smith), University of Illinois at Chicago, Chicago, IL. Supported by the American Occupational Therapy Foundation, the University of Illinois at Chicago Graduate College and the National Institute of Disability and Rehabilitation Research ARRT (grant no. H133P050001). No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Reprint requests to Jessica M. Kramer, PhD, Boston University, Sargent College, Occupational Therapy, 635 Commonwealth Ave (SAR 512), Boston, MA 02215, e-mail: [email protected]. 0003-9993/09/9012-00404$36.00/0 doi:10.1016/j.apmr.2009.07.019

Participants Participants in this study sample were selected from a database that included 502 children. This database represented a convenience sample of children meeting the following inclusion criteria: age 6 to 17 years, completion of the COSA in their primary language, and diagnosed with a disability or receiving occupational therapy. Children who responded to

List of Abbreviations AIC BIC COSA DIF

Akaike information criterion Bayesian information criterion Child Occupational Self Assessment differential item functioning

Arch Phys Med Rehabil Vol 90, December 2009

2048

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer

Table 1: Demographics for Competence and Importance Samples

Children’s Demographic Information

Sex Boys Girls Ethnicity White Black Hispanic Other (eg, multiracial, Asian) Missing information Major condition category Developmental delay Neurologic Mental health Other (eg, orthopedic, OT only) Missing information Practice setting School (inclusive) setting School (center) setting Other setting (eg, inpatient, outpatient) Country United States UK Other Age (y, mo), mean ⫾ SD

Competence Sample (N⫽407)

Importance Sample (N⫽401)

n

%

n

%

286 121

70.3 29.7

283 118

70.6 29.4

323 32 20

79.4 7.8 4.9

326 26 18

81.3 6.4 4.5

31 1

7.6 0.2

30 1

7.5 0.2

257 77 34

63.1 18.9 8.4

255 75 33

63.6 18.7 8.2

38 1

9.3 0.2

37 1

9.2 0.2

221 88

54.3 21.6

214 88

53.4 21.9

98

24.1

99

24.7

230 56.5 121 29.7 56 13.8 (n⫽403) 11,9.62⫾2,10.46

paper-and-pencil form that provides visual cues (faces and stars) and verbal descriptors for each rating category, (2) a card sort version that uses the same visual cues and descriptors, and (3) a summary form with only verbal descriptors. The COSA was developed in English, but the database included COSAs administered in Italian, Icelandic, German, and British Sign Language. Procedures A university research ethics committee approved this research protocol. Researchers and clinicians using the COSA were invited to contribute children’s COSA responses to a central database. Basic demographic information was obtained by each contributor from the child’s medical, school, or research record. Ninety-eight practitioners and researchers who were occupational and physical therapists contributed data. Each contributor selected the method of administration and provided modifications as specified in the administration manual to meet the individual needs of the child. Analysis Researchers used the Mixed Rasch Model15 to explore the relationship between self-report response patterns and personal and contextual variables. The Mixed Rasch Model combines the benefits of Rasch measurement with latent class theory and identifies different groups, or “classes,” of respondents for

220 54.9 121 30.2 60 15.0 (n⫽397) 11,8.81⫾2,9.29

Abbreviations: OT, occupational therapy; UK, United Kingdom.

fewer than 10 items were not included in the database. Analytical software restrictions limited inclusion in this study to those children with full responses to at least 1 scale. This resulted in a competence and an importance sample for this study (table 1). Differences between children selected for inclusion in this study from the database and those not included were tested using chi-square statistics and 2-sample t tests. In the competence sample, 288 (70.8%) children completed the assessment using the standard COSA paper-and-pencil form, 127 children (31.2%) required modifications to complete the COSA, 62 children (15.2%) used a translation, and 75 (18.4%) children had intellectual disabilities. In the importance sample, 287 (71.6%) children completed the assessment using the standard form, 120 (29.9%) required modifications, 66 children (16.5%) used a translation, and 73 (18.2%) children had an intellectual disability. Instrument: The Child Occupational Self Assessment The COSA consists of 25 items (fig 1) and two 4-point scales: competence and importance. Children age 6 to 17 can use the COSA.12-14 Administrators determine the appropriateness of the COSA by considering a child’s capacity to identify preferences and reflect on personal strengths and challenges, as well as the relevancy of the items to each child’s current situation. The COSA can be administered using (1) a standard Arch Phys Med Rehabil Vol 90, December 2009

Fig 1. COSA items and rating scale categories.

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer

RESULTS Competence Sample Compared with the children included in the competence sample, children not selected for inclusion from the database were older (mean ⫾ SD, 12.8⫾3.2y vs 11.8⫾2.7y; P⬍.01), more likely to be from the United Kingdom (69.5% vs 29.7%; P⬍.01), more likely to be seen in a center school setting (ie, a noninclusive school setting providing educational services only to children and youth with disabilities) (59% vs 21.6%; P⬍.01), and less likely to use a translation (7.4% vs 15.2%; P⬍.05). Fit statistics indicated that 2-class (AIC, 21,859; BIC, 22,088) and 3-class (AIC, 21,731; BIC, 22,075) solutions had similar model fit statistics. Within the 3-class solution, 1 class had 9 items with poor fit to the model. Within the 2-class solution, only 1 item misfit in 1 class. As a result, the 2-class solution was further explored as the best solution.

Rating scale Threshold difficulty

Competence class 1

COSA Items Competence class 2

Rating scale Threshold difficulty

whom the Rasch Rating Scale Model holds. For each latent class, the Mixed Rasch Model estimates a set of Rasch parameters for each item, child, and rating scale.16,17 The model was run using the WINMIRA software,a which does not allow data with missing responses. We believe the Mixed Rasch Model is better suited than a DIF approach to answer this study’s research question. The main reason is that when conducting a DIF analysis, researchers compare differences in item difficulty across manifest (ie, observable) categorical variables, but with the Mixed Rasch Model, researchers can compare differences in relative item difficulty, item fit, and rating scale use across empirically defined latent classes. This provides a more empirical exploratory investigation of how different latent classes of children use and interpret self-report assessments. DIF requires a researcher to identify personal or contextual variables a priori to study differences in item difficulty, while the Mixed Rasch Model divides people into latent classes for which different item difficulties and thresholds are present. This decreases the influence of researcher bias when examining group differences. The Rasch Rating Scale Model provides detailed information regarding children’s use of rating scale categories18 by estimating the relative position of each rating category relative to another along a hypothesized continuum. On the COSA, rating scales were designed to indicate a continuum of less to more competence and importance. Therefore, the transition between “I have a big problem doing this” and “I have a little problem doing this” should measure less competence than the transition between “I have a little problem doing this” and “I do this OK.” This transition is referred to as a threshold. Children use the rating scale in the intended manner when thresholds increase incrementally.18 A 2-step process determined the number of latent classes that represented the best Mixed Rasch Model solution. The first step is to identify statistically the best fitting model (both overall and at the item level); the second step is to ensure that solution provides substantively meaningful results (ie, not statistically optimized in a theoretical vacuum). WINMIRA provides 2 overall model fit statistics: AIC and BIC.16,17 For each of these statistics, a lower number indicated better overall fit. The fit of individual items is assessed using the Q-index. Items with a Q-index between 0.0 and 0.3 had good fit to the Rasch rating scale model.17 A standardized Q-index is also provided, and significance was indicated by probability less than .05 or greater than .95.17 Group differences were explored using chi-square statistics and between-group t tests.

2049

COSA Items Threshold 3: “I do this OK” to “I am really good at doing this” Threshold 2: “I have a little problem doing this” to “I do this OK” Threshold 1: “I have a big problem doing this” to “I have a little problem doing this”

Fig 2. Competence rating scale threshold parameters for Competence class 1 and Competence class 2.

Competence class 1 included 212 children (50.8%). Children in competence class 1 used the competence rating scale as a continuum from less to more competence, as indicated by ordered thresholds that increased incrementally (fig 2). All items had Q-indexes between 0.0 and 0.3, but 1 standardized Q-index indicated a significant misfit for the item, “Make my body do what I want it to do.” Competence class 2 included 195 children (49.2%). Children in competence class 2 did not use the rating scale as intended, and step thresholds decreased incrementally with each rating scale category transition (see fig 2). All items had acceptable fit on the competence scale as indicated by Q-indexes and standardized Q-indexes. Children in competence classes 1 and 2 differed in the relative amount of competence reported for some everyday activities. Competence class 1 found “Get my chores done,” “Think of ways to do things when I have a problem,” “Choose things that I want to do,” and “Make others understand my ideas” more difficult than competence class 2. Conversely, competence class 2 found it more difficult to “Get my homework done” and “Finish my work in class on time” in comparison with other everyday activities. Membership in competence class 1 and 2 varied with uses of modifications, use of a translation, presence of intellectual disability, country of residence, and age (table 2). Additionally, all children in this sample with poor fit to the competence scale in the previous study14 belonged to competence class 2 in this analysis. Class membership did not vary as a function of sex, ethnicity, practice setting, disability condition, or administration method. Arch Phys Med Rehabil Vol 90, December 2009

2050

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer Table 2: Competence Class 1 and 2 Significant Differences on Personal and Contextual Variables* Competence Class 1 (n⫽212) Variable

Use of modifications Translation Intellectual disability Country

Mean age ⫾ SD (y)

Variable Category

Yes No English Other Yes No United States UK Other 12.21⫾2.75

Competence Class 2 (n⫽195)

n (%)

Variable Category

57 (26.9) 155 (73.1) 170 (80.2) 42 (19.8) 30 (14.2) 182 (85.8) 108 (50.94) 63 (29.72) 41 (19.34)

Yes No English Other Yes No United States UK Other 11.37⫾2.69

n (%)

P

70 (35.9) 125 (64.1) 175 (89.7) 20 (10.3) 45 (23.3) 148 (76.7) 122 (62.6) 58 (29.7) 15 (7.7)

⬍.05 ⬍.01 ⬍.05 ⬍.01

⬍.01

*Univariate associations. Abbreviation: UK, United Kingdom.

DISCUSSION The purpose of this study was to determine whether all children in a sample interpreted the items and rating scales of a self-report in a consistent and expected manner, and if not, to Arch Phys Med Rehabil Vol 90, December 2009

identify the differences in the item difficulties, rating scale thresholds, and any personal and contextual variables that distinguished differences between latent classes. Because the database from which this study sample was drawn is not representative of the population of all children with disabilities, results from this study should be interpreted with caution. Further, the children selected for inclusion in this study from the database were significantly different from those not selected on the variables of age, country of residence, practice setting, use of translation, presence of intellectual disability, and use of modifications.

Rating scale Threshold difficulty

Importance class 1

COSA Items Importance class 2

Rating scale Threshold difficulty

Importance Sample Compared with children included in the importance sample, children not selected for inclusion from the database were older (13⫾3y vs 11.7⫾2.8y; P⬍.01), more likely to be from the United Kingdom (65.3% vs 30.2%; P⬍.01), more likely to be seen in a center school setting (55.4% vs 21.9%; P⬍.01), and less likely to use a translation (3% vs 16.5%; P⬍.01). Additionally, they were more likely to have an intellectual disability (27.2% vs 18.3%; P⬍.05) and more likely to use modifications (41.6% vs 29.9%; P⬍.05). Fit statistics indicated that the 2-class (AIC, 22,551; BIC, 22,779) and 3-class (AIC, 22,325; BIC, 22,669) solutions had similar model fit statistics. In the 3-class solution, 1 class had 3 items with poor fit to the model. In the 2-class solution, all items met fit standards. The 2-class solution was further explored as the best solution. Importance class 1 included 179 children (45.6%). Children in importance class 1 used the rating scale as the intended continuum, as indicated by ordered thresholds that increased incrementally (fig 3). All items had acceptable fit to the Rasch model as given by Q-indexes and standardized Q-indexes. Importance class 2 included 222 children (54.4%). Importance class 2 did not use the rating scale as intended; thresholds were disordered (see fig 3). All items had acceptable fit to the importance scale as given by Q-indexes and standardized Qindexes. Importance classes 1 and 2 differed in the relative importance reported for some everyday activities. Importance class 1 reported that items pertaining to personal pleasure, such as “Have enough time to do things I like” and “Choose things that I want to do,” were less important than school tasks. Conversely, importance class 2 reported that items pertaining to personal pleasure were more important than school tasks. Membership in importance class 1 or 2 varied with practice setting, use of a translation, presence of intellectual disability, disability condition, country of residence, and age (table 3). All children in this sample who had inconsistent response patterns on the importance scale in a previous study14 belonged to importance class 2 in this analysis. Class membership did not vary as a function of sex, ethnicity, or administration method.

COSA Items Threshold 3: “Really important to me” to “Most important of all to me” Threshold 2: “Important to me” to “Really important to me” Threshold 1: “Not really important to me” to “Important to me”

Fig 3. Importance rating scale threshold parameters for Importance class 1 and Importance class 2.

2051

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer Table 3: Importance Class 1 and 2 Significant Differences on Personal and Contextual Variables* Importance Class 1 (n⫽179) Variable

Practice setting

Translation Intellectual disability Disability

Country

Mean age ⫾ SD (y)

Variable Category

School (inclusive) School (center) Other English Other Yes No DD Neuro MH Other United States UK Other 12.18⫾2.68

Importance Class 2 (n⫽222)

n (%)

Variable Category

100 (55.9) 28 (15.6) 51 (28.5) 138 (77.1) 41 (22.9) 24 (13.5) 154 (86.5) 110 (61.5) 38 (21.2) 9 (5.0) 22 (12.3) 91 (50.8) 48 (26.8) 40 (22.4)

School (inclusive) School (center) Other English Other Yes No DD Neuro MH Other United States UK Other 11.38⫾2.80

n (%)

P

114 (51.4) 60 (27.0) 48 (21.6) 197 (88.7) 25 (11.3) 49 (22.2) 172 (77.8) 145 (65.6) 37 (16.7) 24 (10.9) 15 (6.8) 129 (58.1) 73 (32.9) 20 (9.0)

⬍.05

⬍.01 ⬍.05 ⬍.05

⬍.01

⬍.01

*Univariate associations. Abbreviations: DD, developmental delay; MH, mental health; Neuro, neurologic condition; UK, United Kingdom.

In this sample, almost 50% of children with disabilities used the COSA competence and importance rating scales as intended, as 4-point scales; the remainder used the scales in an unintended manner. The placement of the third rating scale threshold indicates this second group of children was most likely to respond using the highest rating scale category, even when reporting lower overall competence or importance for everyday activities. Further, the proximity of the first and second thresholds suggests that these children were unable to make meaningful distinctions between the lower 3 rating categories. Therefore, one could describe this group’s use of the scales as reversed 2-point scales. However, across classes on both scales, all but 1 item had acceptable fit to the Rasch model. This demonstrates that even when children in this sample used the rating scale in an unintended manner, items were interpreted in a consistent manner.19 While several personal variables were associated with unintended rating scale use in this sample, younger age and the presence of an intellectual disability were variables common across both scales. In this study, rating scale categories were not meaningfully differentiated by children in Competence and Importance class 2, who were significantly younger than children in class 1. These findings are aligned with other research that suggests older children are better able to identify nuances in their abilities and feelings and then translate those differences to a rating scale.20-22 The visual images accompanying the rating scales were intended to facilitate children’s understanding of the differences between categories. However, it is possible that for the importance scale, the use of stars to depict both the positive and negative ends of the rating scale continuum confused children. Future research should explore whether other symbols better support younger children’s interpretation of both rating scales. In this sample, children with intellectual disabilities were also more likely to use the rating scale in an unintended manner. While some researchers propose that difficulty synthesizing information makes self-reports unreliable for persons with intellectual disabilities,23 other research suggests that children with intellectual disabilities can reliably use picture-based self-report scales when rating decisions are presented as a series of two 2-point scales.24-26 Future research should explore

the impact that a similarly modified administration has on children with intellectual disabilities’ use of the COSA. The mixed Rasch model identified that the children’s relative competence and importance for some everyday activities differed across groups. The differences in reported competence for everyday activities such as solving problems and making others understand one’s ideas may reflect age-related expectations, because children in competence class 1 were older. Older children may find these activities more difficult to perform in a way that meets adults’ expectations. The differences in the reported importance for school tasks versus personal satisfaction may be related to the increased likelihood that children in importance class 2 received services in noninclusive school settings. These children may report more value for exercising personal choice and control because they have fewer opportunities to do so within more restrictive school environments. Additionally, children in importance class 1 were older, and the increased importance reported for school-related activities may reflect the increased expectations placed on older students to achieve in school. This study revealed why, in previous research,14 some children had poor fit to the Rasch model. These children were using the rating scale in a consistent, albeit unintended, manner: as a reversed 2-point rating scale. The conclusions drawn from these differing interpretations are not inconsequential; inconsistent use of a rating scale as indicated by poor fit assumes an inability to engage in the self-report process, while use of the rating scale as a 2-point scale suggests that children can engage in a modified self-report process. The association between translation and rating scale use in this sample points to a potential problem in the translated versions of the COSA. Translations may have changed the inherent meaning of the rating scale category descriptors and therefore changed the manner in which children responded to those categories. Although COSA translation procedures sought to maintain the meaning of the original English rating categories, translation procedures varied by country, and separate validation procedures may be required if the preferred translation leads to conceptual differences in rating category descriptors. Arch Phys Med Rehabil Vol 90, December 2009

2052

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer

Study Limitations This analysis used a nonrepresentative convenience sample that administered the COSA in a variety of ways. This lack of standardization across respondents decreases the certainty of the conclusions that can be drawn from this analysis. Further, the demographics available for analysis limit the conclusions that can be drawn regarding group differences. Finally, selection bias may have increased the likelihood of finding significant differences in personal or contextual variables across classes, and limits the extent to which conclusions can be drawn. For example, therapists contributing data from several residential schools in the United Kingdom indicated that older students in these settings did not have homework because of the highly regimented schedule. Many of these children were not included in this study; it is unknown whether these children represent a third latent class of respondents. Future research should replicate this analysis with a representative sample with complete data to support the current findings. Interpreting Responses: Implications for Practice These findings suggest that professionals should support children to engage in self-reports to the best extent possible. The literature suggests the following strategies to facilitate reflection and valid self-reports: photographs and drawings, support for reading and writing, and nonleading questions.27-29 When interpreting the self-reports of children with disabilities, particularly those who are younger than 12 years or who have intellectual disabilities, professionals may need to shift their “scaling” of the rating scale. For these children, the use of any rating category except the highest may indicate an awareness of a challenge. As an alternative, professionals could present this group of children with a modified 2-point response scale to enhance the reliability of the self-reports. However, in this study, some younger children and children with intellectual disabilities used the COSA rating scales in the intended manner. Therefore, the decision to provide younger children and those with intellectual disabilities with a 2-point scale should be based on each child’s ability to discriminate between rating categories. CONCLUSIONS This study identified groups of children who respond differently to a self-report of competence and importance for everyday activities. Because the study sample is not representative of the population of all children with disabilities, results from this study should be interpreted with caution. In this sample, about 50% of children used the rating scales as the intended 4-point scales. About 50% used the rating scales as reversed 2-point scales; these children were more likely to be younger and to have intellectual disabilities. However, all items but 1 had acceptable fit to the Rasch model for all groups, suggesting that the children in this sample interpreted items in a consistent manner. Acknowledgments: We thank all the researchers and clinicians who used the COSA and shared responses for analysis in this study, especially clinicians at Occupational and Physical Therapy Department, Fairfax County Public Schools, Maryland; Occupational Therapy Department, Central Manchester and Manchester Children’s University Hospital, United Kingdom; Alderwasley/St. John’s/Callow Park College Occupational Therapy Department, United Kingdom; Occupational Therapy Department, Kennedy Krieger Institute, Maryland; Croydon Primary Care National Health Service Trust, United Kingdom; Southwest London and St. George Mental Health NHS Trust, United Kingdom; Children’s Center Worthing Hospital, United Kingdom; Carilion Clinic Children’s Hospital, Virginia; University of Arch Phys Med Rehabil Vol 90, December 2009

Illinois Medical Center at Chicago, IL; and researchers Sara Harney, MS, OTR/L; Dr. Jane O’Brien, University of New England; Maike Wolfe and Ines Pätzold, Germany; Hronn Birgisdottir, Kristjana Olafsdttr, and Erla B. Sveinbjornsd, Iceland; and Sabrina Schwarz and Lietta Santinelli, Switzerland. References 1. Dickinson HO, Parkinson KN, Ravens-Sieberer U, et al. Selfreported quality of life of 8 –12-year-old children with cerebral palsy: a cross-sectional European study. Lancet 2007;396:2171-8. 2. Engelbert RH, Gulmans VA, Uiterwaal CS, Helders PJ. Osteogenesis imperfecta in childhood: perceived competence in relation to impairment and disability. Arch Phys Med Rehabil 2001;82: 943-8. 3. Law M, King G, King S, et al. Patterns of participation in recreational and leisure activities among children with complex physical disabilities. Dev Med Child Neurol 2006;48:337-42. 4. Shields N, Loy Y, Murdoch A, Taylor NF, Dodd KJ. Self-concept of children with cerebral palsy compared with that of children without impairment. Dev Med Child Neurol 2007;49:350-4. 5. Steiner SJ, Pfefferkorn MD, Fitzgerald JF. Patient-reported symptoms after pediatric outpatient colonoscopy or flexible sigmoidoscopy under general anesthesia. J Pediatr Gastr Nutr 2006;43: 483-6. 6. Sirken MG, Herrmann DJ, Schechter S, Schwarz N, Tanur JM, Tourangeau R, editors. Cognition and survey research. New York: Wiley-Interscience Publication; 1999. 7. Harter S. The construction of the self: a developmental perspective. New York: Guilford Pr; 1999. 8. Wehmeyer ML, Garner NW. The impact of personal characteristics of people with intellectual and developmental disabilities on self-determination and autonomous functioning. J Appl Res Intellect 2003;16:255-65. 9. Keller J, Kafkes A, Basu S, Federico J, Kielhofner G. The Child Occupational Self Assessment (COSA). Version 2.1. Chicago: MOHO Clearinghouse, Department of Occupational Therapy, College of Applied Health Sciences, University of Illinois at Chicago; 2005. 10. World Health Organization. International classification of functioning, disability, and health. Geneva: World Health Organization; 2001. 11. American Occupational Therapy Association. Occupational therapy practice framework: domain and process. 2nd ed. Am J Occup Ther 2008;62:625-83. 12. Keller J, Kafkes A, Kielhofner G. Psychometric characteristics of the Child Occupational Self Assessment (COSA) part one: an initial examination of psychometric properties. Scand J Occup Ther 2005;12:118-27. 13. Keller J, Kielhofner G. Psychometric characteristics of the Child Occupational Self Assessment (COSA) part two: refining the psychometric properties. Scand J Occup Ther 2005;12:147-58. 14. Kramer J, Kielhofner G, Smith EV Jr. Validity evidence for the Child Occupational Self Assessment (COSA). Am J Occup Ther. In press. 15. Rost J. Rasch models in latent classes: an integration of two approaches to item analysis. Appl Psychol Meas 1990;14:271-82. 16. Schmidt KM. Using the mixed Rasch model to discover latent classes of cognitive self-efficacy. Paper presented at the 11th Annual International Objective Measurement Workshop, April 6 –7, 2002. New Orleans: International Objective Measurement Workshop (IOMW); 2002. 17. Wagner-Menghin MM. The mixed-Rasch model: an example for analyzing the meaning of response latencies in a personality questionnaire. J Appl Meas 2006;7:225-37. 18. Linacre JM. Optimizing rating scale category effectiveness. In: Smith EV Jr, Smith RM, editors. Introduction to Rasch measurement. Maple Grove: JAM Pr; 2004. p 258-78.

RATING SCALE USE BY CHILDREN WITH DISABILITIES, Kramer

19. Hong S, Min S-Y. Mixed Rasch modeling of the Self-Rating Depression Scale. Educ Psychol Meas 2007;67:280-99. 20. Chambers CT, Johnston C. Developmental differences in children’s use of rating scales. J Pediatr Psychol 2002;27:27-36. 21. Harter S, Pike R. The pictorial scale of perceived competence and social acceptance for young children. Child Dev 1984;55:1969-82. 22. Shields BJ, Palermo TM, Powers JD, Fernandez SA, Smith GA. The role of developmental and contextual factors in predicting children’s use of a visual analogue scale. Child Health Care 2005;34:273-87. 23. Harter S. Processes in the formation, maintenance, and enhancement of the self-concept. In: Suls J, Greenwald A, editors. Psychological perspective on the self. Hills-Dale: Lawrence Erlbaum; 1986. p 137-81. 24. Elias C, Vermeer A, Hart H. Measurement of perceived competence in Dutch children with mild intellectual disabilities. J Intell Disabil Res 2005;49:288-95.

2053

25. Glenn S, Cunningham C. Evaluation of self by young people with Down syndrome. Int J Disabil Dev Educ 2001;48:163-77. 26. Vermeer A, Lijnse M, Lindhout M. Measuring perceived competence and social acceptance in individuals with intellectual disabilities. Eur J Spec Needs Educ 2004;19:283-300. 27. Brewster SJ. Putting words into their mouths? interviewing people with learning disabilities and little/no speech. Br J Learn Disabil 2004;32:166-9. 28. Kelly B. Methodological issues for qualitative research with learning disabled children. Int J Soc Res Methodol 2007;10:21-35. 29. Milne R, Bull R. Does the cognitive interview help children to resist the effects of suggestive questioning? Legal Criminol Psychol 2003;8:21-38. Supplier a. WINMIRA Software; WINMIRA, Kiel, Germany, e-mail: [email protected].

Arch Phys Med Rehabil Vol 90, December 2009